EWALD

References: Ewald1921, Darden1993, Essmann1995, Toukmaji1996, Laino2008

Ewald parameters controlling electrostatic only for CLASSICAL MM. [Edit on GitHub]

Subsections

Keywords

Keyword descriptions

ALPHA: real = 3.50000000E-001 [angstrom^-1]

Usage: alpha .30

alpha parameter associated with Ewald (EWALD|PME|SPME). Recommended for small systems is alpha = 3.5 / r_cut. For Tight-binding application a recommended value is alpha = 1.0. Tuning alpha, r_cut and gmax is needed to obtain O(N**1.5) scaling for ewald. [Edit on GitHub]

EPSILON: real = 1.00000000E-006

Usage: epsilon 1e-6

tolerance of gaussians for fft interpolation (PME only) [Edit on GitHub]

EWALD_ACCURACY: real = 1.00000000E-006 [hartree]

Usage: EWALD_ACCURACY {real}

Expected accuracy in the Ewald sum. This number affects only the calculation of the cutoff for the real-space term of the ewald summation (EWALD|PME|SPME) as well as the construction of the neighbor lists (if the cutoff for non-bonded terms is smaller than the value employed to compute the EWALD real-space term). This keyword has no effect on the reciprocal space term (which can be tuned independently). [Edit on GitHub]

EWALD_TYPE: enum = EWALD

Usage: EWALD_TYPE (NONE|EWALD|PME|SPME)

Valid values:

  • NONE NONE standard real-space coulomb potential is computed together with the non-bonded contributions

  • EWALD EWALD is the standard non-fft based ewald

  • PME PME is the particle mesh using fft interpolation

  • SPME SPME is the smooth particle mesh using beta-Euler splines (recommended)

References: Ewald1921, Essmann1995, Darden1993

The type of ewald you want to perform. [Edit on GitHub]

GMAX: integer

Usage: gmax 25 25 25

number of grid points (SPME and EWALD). If a single number is specified, the same number of points is used for all three directions on the grid. If three numbers are given, each direction can have a different number of points. The number of points needs to be FFTable (which depends on the library used) and odd for EWALD. The optimal number depends e.g. on alpha and the size of the cell. 1 point per Angstrom is common. [Edit on GitHub]

NS_MAX: integer = 11

Usage: ns_max 11

number of grid points on small mesh (PME only), should be odd. [Edit on GitHub]

O_SPLINE: integer = 6

Usage: o_spline 6

order of the beta-Euler spline (SPME only) [Edit on GitHub]

RCUT: real = [angstrom]

Usage: RCUT 5.0

Explicitly provide the real-space cutoff of the ewald summation (EWALD|PME|SPME). This value is ignored in Tight-binding applications (rcut from basis overlap is used). If present, overwrites the estimate of EWALD_ACCURACY and may affect the construction of the neighbor lists for non-bonded terms (in FIST), if the value specified is larger than the cutoff for non-bonded interactions. [Edit on GitHub]